D1 · Cloud Concepts

The single most important principle for understanding cloud security. It defines exactly who's responsible for what at each layer of the technology stack — and it changes depending on the service model.

The fundamental concept

In the cloud, security and management aren't the exclusive responsibility of any single party. Microsoft and you share responsibilities — the distribution depends on the service model you use.

This model exists because in a physical datacenter, you control absolutely everything. When you move workloads to the cloud, you hand off certain control to the provider in exchange for them managing those layers. The model defines exactly where the line is.

Why this concept exists (and why it matters)

In 2021, 48% of cloud security breaches were caused by customer misconfiguration — not Microsoft vulnerabilities. In every one of those cases, the customer wrongly assumed "Microsoft handles everything".

Understanding this model is the difference between a secure architecture and an average $4M data breach.

Complete responsibility matrix

Every layer of the technology stack, and who's responsible for it depending on the service model.

Stack layerOn-PremisesIaaSPaaSSaaS
Physical facilitiesCustomerMicrosoftMicrosoftMicrosoft
Hardware (servers, storage, switches)CustomerMicrosoftMicrosoftMicrosoft
Physical networkCustomerMicrosoftMicrosoftMicrosoft
Hypervisor / VirtualizationCustomerMicrosoftMicrosoftMicrosoft
Operating SystemCustomerCustomerMicrosoftMicrosoft
Middleware / RuntimeCustomerCustomerMicrosoftMicrosoft
ApplicationCustomerCustomerCustomerMicrosoft
Application dataCustomerCustomerCustomerShared
Endpoints (user devices)CustomerCustomerCustomerCustomer
Identity and access (IAM)CustomerCustomerSharedShared
Network and firewall (config)CustomerSharedSharedMicrosoft
MicrosoftFull responsibility of Microsoft
CustomerFull responsibility of the customer
SharedBoth share responsibility

What is ALWAYS each party's job

No matter the service model (IaaS, PaaS, or SaaS), these responsibilities never change.

Microsoft is ALWAYS responsible for:

  • Physical datacenter: Buildings, electricity, cooling, perimeter security, natural disasters
  • Hardware: Servers, racks, switches, physical storage — purchase, installation, replacement
  • Global physical network: Fiber cables, routers, Azure backbone network
  • Hypervisor: Hyper-V and the virtualization stack. Isolation between tenants
  • Platform availability: Azure SLAs. If Azure fails, they compensate per the
  • Physical security: Physical access control to datacenters. Only certified Microsoft employees get in

You're ALWAYS responsible for:

  • Your data: Encryption, backup, classification of sensitive data, GDPR/HIPAA compliance for the data you store
  • Identity and access: Who can access what. Account management, MFA, least privilege
  • User endpoints: The devices your employees use to access Azure services
  • Managing external access: Exposed APIs, applications with public access, API keys
  • Applicable legal compliance: Knowing which laws apply to your data and ensuring your configuration meets them
  • Backing up critical data: Even though Azure offers backups, the strategy and validation are your responsibility

— Maximum customer responsibility

In IaaS (), you control everything from the OS up. That means more power and more responsibility.

Example: Azure Virtual Machine with Windows Server 2022

Microsoft manages:

  • The physical server your VM runs on
  • The Hyper-V hypervisor
  • The datacenter's physical network
  • The physical storage of the disks
  • The availability of the physical host

You manage:

  • Windows Server OS (monthly patches)
  • Antivirus and antimalware
  • IIS, .NET, SQL Server if you install them
  • Your application and its code
  • Windows firewall (NSG + OS)
  • RDP/SSH access and who can connect
  • VM backups (even though Azure Backup exists)
  • Critical security updates

Real-world IaaS breach scenario

Company X has 200 VMs in Azure IaaS. The sysadmin doesn't apply the April Windows patch (critical CVE). An attacker exploits the vulnerability, gains access to 50 VMs, exfiltrates customer data.

Who's responsible? The customer. Microsoft delivered a functioning OS. The customer was responsible for patching it and didn't. Microsoft does NOT have access to the VM to patch it without your consent.

— Divided responsibility, less work

In PaaS (Azure App Service, Azure SQL Database), Microsoft manages the OS and runtime. You only worry about your code and data.

Example: Azure App Service (Web App)

Microsoft manages:

  • The server OS (Windows/Linux)
  • Automatic OS patches
  • .NET, Node, Python, Java runtime
  • IIS / underlying web server
  • Automatic redundancy across AZs
  • SSL/TLS termination
  • Basic DDoS at the platform level
  • Automatic platform scaling

You manage:

  • Your code and its dependencies
  • Environment variables and config
  • Who can deploy ()
  • Connection strings and secrets
  • Application data
  • Logging and monitoring of your app
  • Authentication of your app's users
  • Custom domain and your own TLS certificates

"Shared" responsibility in PaaS: IAM and Network

Network and firewall

Microsoft handles the physical network. You configure the rules: which IPs can access (restrictions), VNet integration, Private Endpoints. If you leave public access open by mistake, it's your responsibility.

Identity and access

Microsoft provides Entra ID. You decide who has which roles in your App Service. A developer with an Owner role they shouldn't have = your mistake, not Microsoft's.

— Minimum responsibility, maximum delegation

In SaaS (Microsoft 365, Teams, SharePoint Online), Microsoft manages almost everything. But your responsibilities remain critical.

Example: Microsoft 365 (Exchange Online + SharePoint)

Microsoft manages:

  • All the infrastructure
  • App updates (automatic)
  • Availability and SLA
  • Platform security
  • Built-in anti-spam and antivirus
  • Infrastructure backups
  • Platform compliance

You're still responsible for:

  • Who has access (licenses and permissions)
  • Data your users upload
  • Data retention policies
  • DLP (Data Loss Prevention): preventing data from leaking out
  • User data backup (Microsoft only retains 30-93 days)
  • Configuring MFA for all users
  • Managing former employees (revoking access)

Common critical mistake in SaaS

A company assumes "Microsoft backs up Teams/SharePoint" and doesn't implement Veeam or similar. An employee accidentally deletes 3 years of SharePoint documents. Microsoft only guarantees restoration up to 93 days in SharePoint (and deleted Teams chat may be unrecoverable).

Lesson: In SaaS, you're still responsible for the business continuity of YOUR data. Microsoft protects its platform, not your specific data.

What happens when you fail your part

The most common — and costly — mistakes happen when customers assume Microsoft covers something that's actually the customer's responsibility.

Storage account with public access enabled

IaaS/PaaSResponsible: Customer

Azure Storage lets you enable "Allow public access" at the account level. If you turn this on + don't have logging, you expose data publicly. Microsoft doesn't prevent it — it's a legitimate option. You're responsible for the configuration.

Capital One (2019): $80M fine + 140M accounts leaked due to S3 misconfiguration.

Admin passwords without MFA

AllResponsible: Customer

Microsoft Entra ID can require MFA — if you configure it. If you create admin accounts without MFA and a weak password, Entra ID correctly authenticates the attacker. Microsoft authenticated correctly; you didn't turn on MFA.

99.9% of compromised-account attacks are prevented by MFA (Microsoft Security Report 2023).

VM with RDP exposed to the internet (port 3389)

IaaSResponsible: Customer

NSGs allow everything by default if you don't configure them. A VM with RDP/SSH open to the world is a direct brute-force target. Microsoft delivered the VM. You configured (or failed to configure) the NSG.

Microsoft blocks 300M+ fraudulent authentication attempts daily — many against VMs with open ports.

Excessive permissions (least privilege violation)

AllResponsible: Customer

A developer with a "Contributor" role can accidentally delete an entire resource group in production. Azure RBAC allows Contributor. You decide who has which role. Microsoft doesn't know whether that developer should have that permission or not.

Human error is the #1 cause of cloud incidents (Gartner 2024).

Shared responsibility and compliance

When you use Azure, you can inherit Microsoft's certifications — but only for the layers Microsoft manages. Your layer still needs compliance.

Azure certifications (that you can inherit)

  • ISO 27001: Information security management
  • SOC 1/2/3: Service organization controls
  • PCI DSS: Payment card data security standard
  • HIPAA: Protected health information (USA)
  • GDPR: EU data regulation
  • FedRAMP: US federal government

What you still have to do for compliance

  • Classify and label your sensitive data
  • Implement encryption of data at rest and in transit
  • Manage access with least-privilege principle
  • Implement logging and access auditing
  • Data retention policies per regulation
  • Impact assessment (DPIA for GDPR)
  • Train employees on data handling

Azure gives you the tools; you have to use them correctly.

Did you get this topic?

Test what you just learned

In the IaaS model, what is the customer responsible for that the cloud provider is not?

Ready to test "Shared responsibility"?

Put what you learned to the test with original AZ-900 exam questions.

Start AZ-900 mock exam

Unlimited mock exams with the Pro plan

Log in to track your progress.